An application of recurrent nets to phone probability estimation

By Anthony J. Robinson. 1994

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Speech & Audio

Citations in the corpus

Pointer Networks (2015, background): “However, their architecture limited them to settings where the inputs and outputs were available at a fixed frame rate (e.g. [4]).”
Deep Learning: Methods and Applications (2014, methodology): “In [304], a separate HMM was used to segment the sequence during training, and the HMM was also used to transform the RNN classification results into label sequences.”
Improved training of end-to-end attention models for speech recognition (2018, methodology): “Conventional speech recognition systems [1] with neural network (NN) based acoustic models using the hybrid hidden Markov models (HMM) / NN approach [2, 3] usually operate on the phone level, given a phonetic pronunciation lexicon (from phones to words).”
Deep Sentence Embedding Using Long Short-Term Memory Networks: Analysis and Application to Information Retrieval (2015, positive): “The RNN is a type of deep neural networks that are “deep” in temporal dimension and it has been used extensively in time sequence modelling [21]–[29].”
A tutorial survey of architectures, algorithms, and applications for deep learning (2014, methodology): “Furthermore, while RNN has early success in phone recognition [114], it was not easy to duplicate due to the intricacy in training, let alone to scale up for larger speech recognition tasks.”
2007 Special Issue: An experimental unification of reservoir computing methods (2007, unknown): “Possible – and actual – applications are manifold and include the learning of context free and context sensitive languages (Gers & Schmidhuber, 2001; Rodriguez, 2001), control and modeling of complex dynamical systems (Suykens, Vandewalle, & De Moor, 1996) and speech recognition (Graves, Eck, Bering”
Sequence-discriminative training of deep neural networks (2013, background): “While computational constraints limited earlier uses of hybrid systems to estimating scaled likelihoods for monophones using a two layered network [2] and recurrent networks [3], recent years have seen a resurgence in their use [4, 5, 6, 7].”
Learning words from sights and sounds: a computational model (2002, background): “This capacity for temporal memory has been shown to effectively model coarticulation effects in speech (Robinson, 1994).”
Discriminative Training for Large-Vocabulary Speech Recognition Using Minimum Classification Error (2007, methodology): “Though far from being a large-scale speech recognition task, the TIMIT phone classification task [30] is a useful benchmark for preliminary experiments.”
Analysis of MLP-Based Hierarchical Phoneme Posterior Probability Estimator (2011, methodology): “G4: Hierarchy Using RNN: Recurrent neural networks (RNNs) can also estimate the phonetic class conditional probabilities [32].”
Sparse connection and pruning in large dynamic artificial neural networks (1997, methodology): “INTRODUCTION It is well-known that artificial neural networks can be successfully used for phoneme recognition problems, e.g., [1,2,3].”
Learning a better representation of speech soundwaves using restricted boltzmann machines (2011, unknown): “6% Recurrent Neural Nets[16] 26.”
Framewise phoneme classification with bidirectional LSTM and other neural network architectures (2005, unknown): “(Robinson, 1994; Bourlard and Morgan, 1994), and a challenging benchmark in sequence processing.”
Connectionist temporal classification: labelling unsegmented sequence data with recurrent neural networks (2006, unknown): “The hybrid system comprised an HMM and a BLSTM network, and was trained using Viterbi-based forced-alignment (Robinson, 1994).”
Bayesian methods for neural networks (2000, background): “In addition, they have been successfully applied to many complex problems, including speech recognition (Robinson, 1994), hand written digit recognition (Le Cun et al.”
An overview of reservoir computing: theory, applications and implementations (2007, background): “The literature describes many possible applications, including the learning of context free and context sensitive languages [41, 11], control and modeling of complex dynamical systems [54] and speech recognition [40, 12].”
A probabilistic framework for segment-based speech recognition (2003, methodology): “…as input to the decoder fit into this framework, including all discrete and continuous observation HMMs, and those using artificial neural networks for classification (Lamel & Gauvain, 1993; Mari, Fohr, & Junqua, 1996; Robinson, 1994; Young & Woodland, 1994; Robinson, Hochberg, & Renals, 1994).”